Skip to main content

Full text of "Habitat Modeling and Protection of American Ginseng (Panax quinquefolius L.) in Great Smoky Mountains National Park"

LIBRARY 

GREAT SMOKY MOUNTAINS 

NATIONAL PARK 



Habitat Modeling and Protection 
of American Ginseng 

(Panax quinquefolius L.) 




in Great Smoky 
Mountains 
National Park 




National Park Service 

107 Park Headquarters Road, Gatlinburg, Tennessee 37738 

December, 1999 



Habitat Modeling and Protection 

of American Ginseng (Panax quinquefolius L.) 

in Great Smoky Mountains National Park 



Janet Rock, J. Hope Hornbeck, Jennifer Tietjen, and Erica Choberka 



U.S. National Park Service 

Great Smoky Mountains National Park 

107 Park Headquarters Road 

Gatlinburg, Tennessee 37738 



December 1999 



CONFIDENTIALITY OF INFORMATION 

"Information concerning the nature and specific location of a National Park 
System resource which is endangered, threatened, rare, or commercially valuable 
. . . may be withheld from the public in response to a request under section 552 of 
Title 5 United States Code ...." Sec. 207. Confidentiality of Information. 
National Parks Omnibus Management Act of 1998. 



This report printed on 100% recycled paper. 



ACKNOWLEDGMENTS 

This project could not have been accomplished without the expertise, and technical assistance of 
John Boetsch and Dr. Frank van Manen. Stephanie Wilds, Jim Corbin, Dr. Peter White, and Dr. 
Daniel Gagnon also contributed to the study's success in model development. Thank you to the 
field technicians and volunteers who put in long hours of hard work in the summers of 1997 and 
1998: Jillian Archer, Beth Buchanan, Chandra Vostral, Eric Nielsen, Jennifer Dean, Mark 
Whited, Jeff Merrick, Daniel Noon, Luke Bruner, Mike Keyes, Brian Yahn and Thomas Moore. 
Additional assistance was provided by Will Blozan, Bob Dellinger, Keith Langdon, and the 
following Great Smoky Mountains National Park rangers: Walt West, Richard Jenkins, John 
Garrison, Lamon Brown, Paul Williams, Al Voner, Ken Davis, Lora Peppers, Steve Spanyer, and 
Mike Farley. And special thanks to Meryl Rose and Don DeFoe for excellent editorial 
comments. Finally, Steve Kemp, Great Smoky Mountains Natural History Association, 
generously designed and produced our report cover. 



ABSTRACT 

A two-year study was initiated to determine the status of American ginseng {Panax 
quinquefolius), in Great Smoky Mountains National Park (GSMNP). The objectives of this 
study were to: 

(1) create a ginseng habitat model, 

(2) estimate the abundance of ginseng in the Park, 

(3) describe ginseng habitat(s) and evaluate species associations, 

(4) locate and mark existing ginseng populations as a poaching deterrent, and 

(5) assess any other pressures on the species. 

A predictive distribution model was developed using a multivariate modeling technique, 
Mahalanobis distance, to identify ginseng habitat in GSMNP. Ninety-eight plots were visited 
during the 1998 field season to refine the model's predictive ability. 

Data suggest that ginseng occupies a wider variety of habitats than previously believed. Our 
hypothesis that ginseng most frequently occurs in rich cove forests was supported by the 
predictive model. However, because this habitat type is targeted for illegal harvest, ginseng can 
often be found in greater abundance in atypical habitats that are less likely to be detected by 
poachers. 

Ordination analyses of environmental and vegetation data were consistent with previous habitat 
studies. Soil analyses revealed significant relationships between pH, zinc, calcium to 
magnesium ratio, and ginseng distributions. Parallel analyses of ginseng and its associated plant 
species suggest that the model accurately predicts ginseng habitat irrespective of the confounding 
factors of illegal harvest. 

To allow identification of illegally harvested ginseng roots from the Park, Park staff located and 
marked roots of wild ginseng plants throughout GSMNP. Further conservation efforts are 
recommended. 

The study was conducted and reported as two separate projects: 

Parti: Habitat Modeling 
Part II: Ginseng Protection 



in 



Table of Contents 



Acknowledgments ii 

Abstract iii 

List of Tables and Figures vi 

Part I: Habitat Modeling 1 

Introduction 1 

Background 1 

The Species 3 

Methods and Materials 5 

Study Area 5 

Habitat Model Development . 6 

Field Validation 6 

Field Methods 6 

Plot Location 6 

Plot Design 6 

Data Collection 7 

Data Analysis 8 

Chi-square and Pearson Correlation Analyses 8 

Detrended Correspondence Analysis (DC A) 8 

Canonical Correspondence Analysis (CCA) 9 

Mahalanobis Distance (D 2 ) Cutoff 9 

Abundance Estimate 9 

Results 10 

Habitat Model Validation 10 

Chi-square Analysis 11 

Pearson Product Moment Correlation 11 

Regression Analyses 12 

Mahalanobis Distance (D 2 ) Cutoff 12 

Ginseng Habitat Description 13 

Plant Species Associations 13 

Canopy Layer 13 

Sub-canopy Layer 13 

Herbaceous Layer 13 

Soil Analysis 14 

Detrended Correspondence Analysis (DC A) 14 

Canonical Correspondence Analysis (CCA) 17 

Distribution Map 18 

Abundance Estimate 18 

Discussion 19 

Ginseng Habitat Description 19 

Associated Plant Species 19 

Large-scale Patterns 19 

Small-scale Patterns 20 

Microtopography and Soil Factors 21 

Abundance Estimate 21 



IV 



Digitized by the Internet Archive 

in 2012 with funding from 

LYRASIS Members and Sloan Foundation 



http://archive.org/details/habitatmodelingpOOrock 



Summary 22 

Recommendations 23 

Part II: Protection 24 

Introduction 24 

Illegal Harvest from GSMNP 24 

Ginseng in GSMNP 26 

Methods and Materials 27 

Protection Techniques 27 

Data Collection 27 

Size Class and Age Distributions 28 

Results 29 

Habitat Data 29 

Associated Plant Species Occurrence 29 

Size Class Distribution 29 

Age Distribution 30 

Seized Roots and Poaching Evidence 31 

Discussion 32 

Literature Cited 33 

Appendices 36 

Appendix 1 : GIS variables used to calculate Mahalanobis 

Distance (D 2 ) 37 

Appendix 2 : Ginseng model validation data sheet 39 

Appendix 3: Species list Detrended Correspondence 

Analysis 40 

Appendix 4: Ginseng Protection data sheet 41 



List of Tables and Figures 

Tables Page 

Table 1 : Ginseng occurrence and abundance in validation plots (n = 98) 10 

Table 2: Chi-square analysis of ginseng presence 11 

Table 3: Pearson correlation coefficients for ginseng occurrence, abundance, 

and associates (n = 98) 14 

Table 4: Pearson correlation coefficients for soil components (n = 78) 14 

Table 5 : Ordination results for CCA of canopy layer dominants and 

environmental variables (n = 73) 17 

Table 6: Summary of amount of illegally harvested ginseng (number of roots 

and undried weight) seized in Great Smoky Mountains National Park, 

1991-1996 24 

Table 7: Average number of roots per dry pound and average price per dried 

Pound, 1991-1999 25 



Figures 



Figure 1 : Distribution of ginseng in North America 3 

Figure 2: Morphology and habit of American ginseng {Panax quinquefolius) ... 4 

Figure 3 : Plot design showing transects, quadrats, and nested sub-plots 7 

Figure 4 : Relationship of ginseng occurrence with D" and class 10 

Figure 5: Relationship of ginseng abundance with D 2 and class 11 

Figure 6: Classification of validation results (n = 98) 12 

Figure 7: DCA of validation plots (n = 96) 15 

Figure 8: DCA of validation plots with ginseng (n = 15) 15 

Figure 9 : DCA of ginseng abundance and soil components 16 

Figure 1 0: CCA plot ordination of dominant canopy species (n = 73) 17 

Figure 1 1 : Size class distribution of ginseng plants encountered in 1998 

(n = 501) 30 

Figure 12: Age distribution of naturally occurring ginseng plants encountered 

during the field season efforts (n = 80) 30 

Figure 1 3 : Age distribution of ginseng plants seized in 1 997 and 1 998 31 



VI 



Part I: HABITAT MODELING 
Introduction 

Part I: 
HABITAT MODELING 

INTRODUCTION 

Background 

American ginseng {Panax quinquefolius) has been illegally collected in Great Smoky Mountains 
National Park (GSMNP) since the Park's establishment (1934). In 1991, Park law enforcement 
rangers began counting the number of roots seized from poachers annually. To date. Park law 
enforcement rangers have intercepted over 10,000 roots. However, this number is believed to be 
only a small percentage of what is actually poached from the Park. Over 9,000 roots have been 
weighed and aged by Park biologists to track the general health of ginseng populations 
throughout the Park, and over 6,500 roots have been replanted for monitoring. Biologists do not 
have an adequate understanding of how many populations of ginseng occur in GSMNP, its 
preferred habitat, or the extent of the poaching problem. 

American ginseng (hereafter "ginseng") is now considered rare or threatened throughout its 
North American range (Nault 1997). Due to the long history of ginseng harvest in North 
America, the population demography of the species is poorly understood. Examination of the 
impact of stochastic environmental events (e.g., forest fire, severe weather episodes such as 
drought) and harvesting on Canadian ginseng populations yielded a minimum viable population 
size of 172 plants (Nantel et al. 1996). This conservative estimate assumes that less than 5% of 
mature plants are harvested annually, seeds are allowed to ripen, and random environmental 
catastrophes will not occur. Additional research in Ontario showed 26% of known ginseng 
populations there became extinct within a 10-year period (1987-1997), of which only seven 
populations qualified as viable (Nault et al. 1998). 

Evidence exists that harvesting pressure on natural populations is increasing. Further, the U. S. 
Forest Service (USFS) reports that collecting of wild roots may be escalating; specifically, 
numbers of legal collecting permits for harvesting wild roots has increased up to 300% over 
previous levels in only a three-year period. GSMNP is flanked by three national forests where 
such activity occurs. 



Part I: HABITAT MODELING 
Introduction 

Throughout the species' range, American ginseng populations most frequently are associated 

with nutrient-rich soils (particularly calcium) with a pH higher than 5.5, and usually occur with 

certain herbaceous species (Nault et al. 1998). Ginseng specialists have consistently observed 

the presence of bloodroot {Sanguinaria canadensis), black cohosh {Cimicifuga spp.), maidenhair 

fern {Adiantum pedatum), and yellow lady's slipper {Cyphpedium pubescens) associated with 

ginseng (J. Corbin, NCDA, pers. comm. and J. Rock, NPS, pers. obs.). Ginseng in the southern 

Appalachians occurs in rich cove forests characterized by high plant diversity, and occupies 

sheltered mesic sites with circumneutral, well-drained, rocky, and deep soils (Whittaker 1956. 

Schafale and Weakley 1990). 

The distribution of rich cove forest in GSMNP has been examined in numerous studies (Cain 
1931, Whittaker 1956, Golden 1981, Parker and Parker 1986, Callaway et al. 1987). These 
studies identified large-scale topographic gradients as the primary determinants of forest type. 
Herbaceous species diversity and richness within cove forests has been regularly studied in 
GSMNP (Bratton 1976, Hicks 1980, Newman and Reddell 1988). These studies identified 
within-stand microtopography, soil gradients, and canopy species associations as the primary 
influences on species distributions. The influence of Eastern hemlock (Tsuga canadensis) has 
also been identified as an important influence on within-stand microtopography and soil 
characteristics (Hicks 1980, Beatty 1984). The distribution of herbaceous species is still poorly 
understood across the mosaic of topographic gradients and within-stand variability of the cove 
forests of GSMNP. 

The remote, rugged terrain of many parts of GSMNP, in combination with the cultural tradition 
and economic incentive for ginseng harvesting in the southern Appalachian region, has resulted 
in serious resource management problems. This two-year study was initiated to evaluate the 
status of ginseng in GSMNP by researching the following questions: 

(1) can a Geographic Information System (GlS)-based predictive habitat model for 
ginseng be created, 

(2) can this model be used to estimate the abundance of ginseng in the Park, 

(3) are other plant species consistently associated with ginseng or ginseng habitat, and 
. (4) do topographic or soil characteristics play a role in ginseng distribution? 



Part I: HABITAT MODELING 
Introduction 

The Species 

American ginseng {Panax quinquefolius) is a long-lived forest herb native from southern Quebec 
through the lake states to central Minnesota and south into several Gulf states (Figure 1). 




Figure 1. Distribution of American ginseng in North America (Argus and White 1989). 



The mature plant has a single, rhizome-borne stem topped with a whorl of three or four leaves, or 
"prongs", each of which is comprised of three to five leaflets (Figure 2). Plant height ranges 
from just six centimeters to nearly one meter tall. Height and leaf number appears to be more a 
factor of nutrient availability and other habitat qualities than of plant age (Lewis and Zenger 
1982). Plants may grow very large in just a few years under optimal environmental conditions, 
or may persist for decades in marginal habitat with very little increase in the size of their roots or 
aboveground stems. The plant produces a single inflorescence bearing tiny white flowers from 
the center of its whorl of leaves. The flowers mature into bright red fruits. Each fruit contains 
one or two seeds, occasionally three in larger plants. Ginseng typically reproduces by seed, 
rarely propagating vegetatively by producing more than one rhizome. Plants must be at least 
several years old in order to flower and produce viable fruit, and ginseng seeds require a two- 



Part I: HABITAT MODELING 
Introduction 

winter stratification period before germination can occur. The combination of these factors 

results in slow rates of population growth (Lewis and Zenger 1982). 




Figure 2. Morphology and habit of American ginseng {Panax quinquefolius) (illustrated by 
Hornbeck 1998). 



Like Asian ginseng {Panax ginseng), the roots of American ginseng are highly valued as a 
medicine and tonic. Although American ginseng is cultivated in the U.S., particularly in the 
north-central states, the wild roots are considerably more valuable in both the North American 
and Asian markets (Robbins 1998). The presumed increased chemical potency and gnarled 
appearance of wild ginseng roots, compared to cultivated roots, are the result of stressful 
growing conditions and slower growth often associated with natural populations. The unique 
shapes of wild roots also are considered a factor in the species' medicinal and mystical value in 
traditional Chinese medicine (Hu 1976). 



Part I: HABITAT MODELING 
Methods and Materials 

METHODS and MATERIALS 

Study Area 

GSMNP covers over 200,000 hectares (800 square miles) along the Tennessee and North 
Carolina border with a perimeter of about 290 kilometers (182 miles). This area contains a wide 
variety of habitats distributed across elevations ranging from 268 to 1,982 meters (880 to 6.684 
feet). Approximately 72% of the Park is made up of hardwood forests, and approximately 45% 
(or roughly 70,000 hectares) of these are rich cove forests (Mackenzie 1991) believed to be the 
forest type most often associated with ginseng occurrence (Lewis and Zenger 1982). GSMNP 
provides the largest protected area for wild American ginseng in the eastern U.S. The southern 
Appalachians, including the Park, are believed to provide the highest quality habitat at the 
southern limit of this species' range. 

Habitat Model Development 

During the 1997 field season, data collection protocols were determined for the 1998 field season 
and ginseng reference points were provided for habitat model development. The Park's East 
District was selected for data collection based on habitat diversity (Mackenzie 1991), a range of 
elevations, and verified populations of ginseng. A field survey of ginseng habitat was conducted 
within the USGS Cove Creek Gap and Luftee Knob quadrangles in the Cataloochee area. The 
predictive model was developed in spring 1998 by John Boetsch (Boetsch and Rock 1999) based 
upon high-quality observations from 1997 data and additional ginseng locations throughout the 
Park (n = 56 positions). The model was applied and tested during the 1998 field season (May - 
August). 

The modeling technique used in this study, Mahalanobis distance, is a unitless. multivariate 
statistic that describes the deviation from an optimal set of conditions (Rao 1952). This 
mathematical function (denoted by D 2 ) was calculated from a set of 13 GIS variables (12 
topographic and a single vegetation type) for each 30 x 30-meter pixel (Appendix 1). 

D was used to describe each pixel's dissimilarity or "distance" from ideal habitat conditions 
based upon the 56 positions. The resulting model is the distribution of the D 2 values across each 
pixel in the Park. We hypothesized that ginseng locations would be correlated with lower D 



Part I: HABITAT MODELING 
Methods and Materials 

values, as the model function predicts that the probability of ginseng occurrence decreases with 

increasing dissimilarity from ideal habitat. Thus, the lower a pixel's D" value, the greater 

probability ginseng would be encountered. 

Field Validation 

Model output was stratified into six validation classes. Classes were related to habitat quality 
and guided the validation effort. Class I represented pixels hypothesized to be most preferred 
habitat and class VI represented the least preferred. Classes were designed so that the area 
within class II was twice that of class I, and so on. One hundred-twenty random positions, 20 in 
each of the six classes, were selected to field-test the model. Stratified sampling was used due to 
the wide range of habitats in GSMNP. Sampling evenly across the entire range of D~ values 
would have resulted in an inordinate number of plots where ginseng is unlikely to be found. 
Therefore, validation efforts were concentrated toward the smaller values of D , which were 
hypothesized to possess optimal ginseng habitat characteristics. In addition, data were collected 
in less-favorable habitats to determine if the model omitted potential habitat and if our 
assumptions were correct. 

Field Methods 

Plot Location 

One hundred twenty random validation points were mapped using Universal Transverse 
Mercator (UTM) coordinates on standard USGS quadrangle maps (7.5 minute, 1 :24,000 scale). 
Coordinates were entered into a global positioning system unit (GPS; PLGR +96, Rockwell 
International, Cedar Rapids, Iowa) for field navigation. Points were located in the field using, 
topographic characteristics, compass bearings, altimeter readings, and an unassisted, military Y- 
code GPS signal from the Department of Defense. The locations were confirmed either by GPS 
(accuracy within 1 5 meters), or by a combination of elevation, aspect, and other topographic 
attributes. 

Plot Design 

Level 1 validation plots were 31.6 x 31.6 meters (1000 m 2 ), which approximated pixel size (900 

m ) in the GIS database. In the field, plots were centered on random points. Two 31.6 meter 



Part I: HABITAT MODELING 
Methods and Materials 

transects defined a plot by orienting one along the contour of the dominant slope and the other 
perpendicular to it, criss-crossing at the plot center. Transects divided the level 1 plot into four 
level 2 quadrats (250 m 2 each). One of these quadrats was randomly selected for division into 
level 3 and 4 nested sub-plots. Orienting from plot center facing out, the nested sub-plots were 
always placed to the end, and to the right, of the transect in each random quadrat (Figure 3). 

3 1 .6 meters 



0> 

E 




Level 1 (1000 m-) 



Level 2 (250 m 2 ) 



Levels (10m') 



Level 4 (1 rrr) 



Figure 3. Plot design showing transects, quadrats, and nested sub-plots. 



Data Collection 

Topographic data (quantitative) and vegetation data (qualitative) were collected throughout the 
1000 m plot (see Data Sheet, Appendix 2). All vascular plant species were listed as well. The 
five dominant species in the canopy (level 1 ), sub-canopy (level 2), and herbaceous layers (levels 
3 and 4) were recorded in order of importance, and cover percentages were estimated for each 
layer to the nearest 10% (100% total coverage). Cover for the sub-canopy layer was estimated at 
level 2 (250 m ) and the herbaceous layer was estimated at level 4 (1 m ) to simplify cover 
estimates for these layers. The presence of four ginseng-associates, bloodroot (Sanguinaria 
canadensis), black cohosh (Cimicifuga spp.), maidenhair fern (Adiantum pedatum), and yellow 
lady's slipper (Cyphpedium pubescens), was recorded. The 1000 m 2 plot, including the three 
sub-plots, was searched for ginseng plants and the smallest sub-plot containing ginseng was 
recorded. 



Part I: HABITAT MODELING 
Methods and Materials 

Soil samples were taken from several locations within each 1000 m plot for analysis. The North 
Carolina Department of Agriculture Agronomic Division (NCDAAD) performed the following 
soil analyses: percent humic matter, acidity, weight per volume of soil, pH, percent basic 
solvency, percent calcium, percent magnesium, ammonium and nitrate content, and potassium, 
manganese, phosphorus, zinc, copper, and sulfur indices. 

Additional data were collected when ginseng was present in the plot including: size class and 
reproductive status of each ginseng plant, any observable damage to the plant, general health of 
the population, and any signs of poaching activity. Concentrations of ginseng encountered while 
traveling to and from the plots were mapped for future study. 

Data Analysis 

Chi-square and Pearson Correlation Analyses 

Variability within and between validation classes was tested with chi-square analyses. 
Relationships between Mahalanobis distance (D 2 ) and ginseng occurrence and abundance, and 
the occurrence of associated plant species and soil components, were identified using Pearson 
correlation coefficients. Ginseng occurrence and abundance data were regressed against 
independent variables (D 2 and quantitative habitat characteristics) using logistic and linear 
regression analyses, respectively. 

Detrended Correspondence Analysis (DCA) 

A comprehensive species-occurrence matrix and environmental variables (elevation, aspect, 
slope, microtopography, slope position, dominant stratum height, canopy gaps, and soil 
components) were ordinated using DCA (Hill and Gauch 1980). DCA generated a two- 
dimensional scatterplot depicting the relative position of the validation plots based upon the 
degree of similarity or difference in plant species composition. General ecological patterns were 
determined from field-validated plot positions by examining individual plot characteristics and 
community associations. Plant species that occurred in less than 5% of the validation plots were 
not included in the analysis. Ordination analysis was conducted using PC-ORD (McCune and 
Medford 1995). 



Part I: HABITAT MODELING 
Methods and Materials 

Canonical Correspondence Analysis (CCA) 

The relationships between five dominant canopy species in each validation plot and 1 1 

environmental variables were examined using CCA (PC-ORD, McCune and Medford 1995). 

These environmental variables are: elevation, aspect, slope, pH, acidity, percent basic solvency, 

cation exchange capacity, phosphorus and zinc indices, percent calcium, and percent magnesium. 

Like DCA, CCA produces a two-dimensional scatterplot depicting the spatial relationship 

between plots based upon the degree of similarity or difference in species composition. Unlike 

DCA, CCA performs a multiple regression of the environmental variables against plot 

relationships. 

Mahalanobis Distance Cutoff 

To refine the model's ability to predict ginseng occurrence, a D 2 cutoff value was identified. The 
percentage of correctly classified ginseng occurrences (X-axis) and correctly classified ginseng 
absences (X-axis) were plotted against the D 2 values (Y-axis). D 2 values at, and below, the point 
at which the two prediction curves intersect, provide a range of ginseng habitats and 
corresponding model values. 

Abundance Estimate 

Ginseng abundance in the Park was estimated from the validation data and the model cutoff. 
The ginseng abundance estimate is the "product" of the number of pixels with a D equal to or 
less than the D 2 cutoff, the saturation rate, and average ginseng abundance. The ginseng habitat 
saturation value was calculated by dividing the number of plots with ginseng by the total number 
of plots visited below the D 2 cutoff. The average ginseng abundance value was determined by 
dividing total ginseng abundance by total number of plots visited for that range. A refined 
abundance estimate also was produced for a broader range of habitats above the D cutoff. 



Part I: HABITAT MODELING 

Results 



RESULTS 



Habitat Model Validation 



Ginseng was present in 15 of the 98 plots sampled (15.3%), and 184 ginseng plants were 
counted. The ranges of Z) 2 values, total D 2 area, pixels per class, number of plots visited for each 
validation class, ginseng abundance, ginseng occurrence, and the occurrence of one or more 
associated plant species are given in Table 1 . 



Table 1. Ginseng occurrence and abundance in validation plots (n = 98). 



Class 


Min. D J 


Max. D' 


Area 


Pixels 


Plots 


Ginseng 


Ginseng 


Associate 








(ha) 


(area/.09) 


visited 


abund. 


occur. 


occur. 


I 


4.170 


6.262 


3,238 


35,982 


16 


7 


2 


13 


II 


6.816 


7.976 


6,476 


71,965 


18 


21 


5 


11 


III 


8.055 


10.010 


12,954 


143,929 


18 


88 


4 


9 


IV 


10.378 


12.716 


25,907 


287,860 


12 


8 


2 


9 


V 


13.845 


17.917 


51,815 


575,718 


15 


60 


2 


1 


VI 


18.384 


62.809 


103,628 


1,151,437 


19 








1 


Totals 






204,018 


2,266,891 


98 


184 


15.3% 


45% 



The relationships between ginseng occurrence and abundance and Mahalanobis distance (D 2 ) are 
illustrated in Figures 4 and 5. Ginseng occurs most often in class II (Figure 4), and ginseng 
abundance is greatest in classes II, III, and V (Figure 5). 



I II III IV v VI 



X! 

~*3 



a. 

OJO 



10 



20 30 40 

Mahalanobis Distance 



50 



60 



70 



Figure 4. Relationship of ginseng occurrence with D and class. 



10 



u 



I II III IV V VI 



Part I: HABITAT MODELING 

Results 



c 
n 

■o 
s 
s 

< 

M 

c 

1/ 



5 18 



58 



48 



38 



28 





• 
• 


4 
• 

• 
• 


• 

• 
» 


• 
• 


• 
• 


• *• • ••«•• ••••• • 















10 



20 30 40 50 

Mahalanobis Distance 



60 



70 



Figure 5. Relationship of ginseng abundance with D and class. 

Chi-square Analysis 

A chi-square test was performed for ginseng occurrence variability across the validation classes. 
The data were analyzed for each class and by grouping classes. The results were not significant; 
however, the P-value was marginally significant for group I-III compared with group IV-VI, 
(Table 2). 

Table 2. Chi-square analysis of ginseng presence. 



Validation Classes 


Chi-square 


Df 


P-value 


I, II, III, IV, V, VI 

I-II, III-IV.V-VI 

I-III, IV-VI 


6.2 
2.8 

3.3 


5 
2 
1 


0.287 
0.246 
0.070 


Pearson Product Mori 


lent Correlation 







Pearson Product Moment Correlation was used to examine relationships between ginseng 
occurrence and ginseng abundance with Mahalanobis distance (D"). The occurrence of ginseng 
was negatively correlated with D (r = -0.194, P < 0.05); that is, ginseng occurrence decreased 
with increasing D~ values. In contrast, ginseng abundance was not correlated with D (r= - 
0.059, P > 0.05). The same analyses for validation plots (n = 15) containing ginseng revealed a 



11 



Part I: HABITAT MODELING 

Results 

positive correlation between increasing ginseng abundance and increasing D 2 values (r = 0.551, 

/><0.05). 

Regression Analyses 

Logistic regression suggests that ginseng occurrence is negatively correlated with lower D~ 
values (n = 98, r = - 0.101, P = 0.078) but only marginally significant. Logistic regression shows 
a significant relationship between D~ and Cimicifuga spp. occurrence (n = 98, r = -0.110, P = 
0.047) and occurrence of one or more associated plant species (n = 98, r = -0.108, P = 0.029). 

Linear regression of ginseng occurrence and D 2 was marginally significant (n = 98, r = -0.006, P 
= 0.056), but linear regression of ginseng abundance and D~, however, was not significant (n = 
98, r - -0.052, P - 0.464). Regression analyses (logistic and linear) of ginseng occurrence and 
soil variables were not significant. However, linear regression of ginseng abundance and pH was 
significant (n = 78, r = 0.341, P = 0.003). 

Mahalanobis Distance CD 2 ) Cutoff 

The model's ability to predict ginseng occurrence was determined by plotting the percentage of 

correctly classified presence and absence against possible D 2 cutoffs (Figure 6). The curve for 



100 



80 



60 



40 



20 

























M 


• 
• •• 




















• 
• 




• 
• 




\ 


i- 
















• 






S ". 


■*. 


\ 


■ 











10 




20 


30 


40 




50 


60 


70 



♦ presence 
■ absence 



Mahalanobis distance (D ) 



Figure 6. Classification of validation results (n = 98). 



12 



Part I: HABITAT MODELING 

Results 

correctly classified presence increased with D , and the curve for correctly classified absence 

decreased with increasing D". Figure 6 shows that 60% of both presence and absence predictions 

are correct for the 8.5 cutoff. 

Ginseng Habitat Description 

Elevations in plots where ginseng occurred (n = 15) ranged from 655 to 1,098 meters (2,160 to 
3,620 feet) with a mean of 900 meters (2,968 feet). Aspect within these plots varied. Slopes 
ranged from 8-36% with a mean of 24.2%. Eight plots had a mid-slope position, four plots were 
situated low on the slope, two plots were high-slope, and one plot occurred at the crest of a ridge. 
Canopy cover ranged from 20-70% with a mean of 61 .3%. 

Plant Species Associations 

Canopy layer 

Dominant canopy species occurring with ginseng included tuliptree {Liriodendron tulipifera), 

red maple {Acer rubrum), and white basswood (Tilia americana var. heterophylla). Additional 

species were yellow buckeye (Aesculus flava), black locust {Robinia pseudoacacia) and 

silverbell {Halesia tetraptera var. monticola). 

Sub-canopy layer 

Dominant sub-canopy species were silverbell, sugar maple {Acer saccharum), red maple and 
striped maple {Acer pensylvanicum). Additional sub-canopy species were yellow buckeye, 
tuliptree, and white basswood. 

Herbaceous layer 

Herbaceous cover ranged from 10-100% with a mean of 52%. Nine of the 15 plots had a cover 
of 50%o or greater. One or more of the ginseng-associated species {Sanguinaria, Cimicifuga, 
Adiantum, and Cypripedium) occurred with ginseng in 13 of the 15 plots. Cimicifuga occurred 
with ginseng in nine plots, Adiantum in eight plots, Sanguinaria in six plots, and Cypripedium in 
two plots. 

Pearson correlation analyses revealed a significant, positive relationship between the occurrence 
of ginseng and Cimicifuga {P < 0.0001), Adiantum {P < 0.0001), Sanguinaria {P < 0.0001), and 



13 



Part I: HABITAT MODELING 

Results 

Cypripedium (P = 0.049). In contrast, ginseng abundance was marginally correlated with 
Cimicifuga (P = 0.06) and Sanguinaha (P = 0.13), and not correlated with Adiantum (P = 0.22) 
or Cypripedium {P = 0.91). Analyses suggest that ginseng occurred in habitats where the 
associated plant species occurred, but was more abundant in habitats where the associated plant 
species were less likely to occur (Table 3). 

Table 3. Pearson correlation coefficients for ginseng occurrence, abundance, and associated 
plant species (n = 98). 





Ginseng 


Ginseng 










Occurrence 


Abundance 


Cimicifuga 


Sanguinaria 


Cypripedium 


Cimicifuga 


0.457*** 


0.190 








Sanguinaria 


0.454*** 


0.154 


0.579*** 






Cypripedium 


0.199* 


0.010 


0.302** 


0.292** 




Adiantum 


0.567*** 


0.125 


0.499*** 


0.447*** 


0.580 



P- values * < 0.05 ** < 0.01 *** < 0.001 

Soil Analysis 

Pearson correlation analyses of soil composition, and ginseng occurrence and abundance, 

indicated strong positive relationships among ginseng occurrence and percent basic solvency 

(BS%), manganese (Mn-I) and sulfur (S-I) indices, and soil pH. Correlation with soil 

components varied between ginseng occurrence and abundance (Table 4). Results of the 

NCAAD tests were incomplete for ammonium and nitrate and, therefore, not included in any 

analyses. 

Table 4. Pearson correlation coefficients for soil components (n = 78). 





Ca:Mg 


BS% 


P-I 


K-I 


Mn-I 


Zn-I 


Cu-I 


S-I 


pH 


Occur. 
P-value 


0.193 

NS 


0.512 

*** 


0.15 

NS 


0.304 

** 


0.416 

*** 


0.231 

* 


0.307 
** 


-0.365 

*** 


0.372 
* ** 


Abund. 

P-value 


0.288 

* 


0.536 

*** 


0.173 

NS 


0.208 

NS 


0.538 

*** 


0.594 

*** 


0.348 

** 


-0.262 
* 


0.125 

NS 



P-value *<0.05 **<0.01 ***< 0.001 

Detrended Correspondence Analysis (DCA) 

Figure 7 displays the results of DCA analyses (n = 96). Figure 8 shows only the validation plots 

in which ginseng was present (n = 15); habitat groupings are outlined to facilitate interpretation. 



14 



Part I: HABITAT MODELING 

Results 

The X-axis appears to represent the interaction of moisture and topographic gradients. The Y- 

axis appears to follow a substrate gradient (e.g., soil composition, parent geology, and/or 

topographic interactions). Plots cluster into three habitat groups: group A represents plots on dry 

ridge tops, group B represents plots in moist, rich coves, and group C represents plots at higher 

elevations. Comprehensive plant species lists for the three habitat groups are given in 

Appendix 3. 



r3 



XI 

en 





Moisture and Topography 
Figure 7. DC A of validation plots (n = 96). 



Figure 8. DCA of validation plots with ginseng 
(n=15). 



Legend for Figures 7 and 8 

A: Dry, ridge-top plots: mid-elevations, low moisture, convex topography and high slope position 

(similar to xeric oak and pine-oak vegetation classifications described by Mackenzie 1993). 

B: Cove hardwood plots: low to middle elevations, high moisture, well-drained, concave topography, 

low-middle slope positions (similar to cove hardwoods and mixed mesic hardwoods classifications 

described by Mackenzie 1993). 

C: High elevation plots: high elevation, high moisture, boggy or seepy, level to convex topography, and 

high slope position (similar to spruce-fir and northern hardwoods classifications described by Mackenzie 

1993). 



Figure 9a shows the abundance of ginseng within each plot. The percentage of Ca, Mg, and BS 
of each plot is shown in Figures 9 b-d. Figures 9e-f give the values of P-I and Mn-I. Analyses 
of the soil data suggest that five of the soil components cluster similarly as the habitat groups. 
Plots located near the bottom of the Y-axis and the center of the X-axis had high percentages of 
Mg, Ca, BS, and high values for Mn-I, and P-I. Plots with ginseng were also found in this 
position. 



15 



Part I: HABITAT MODELING 

Results 



a. Ginseng abundance in habitat groupings b. Ca% 




c. Mg% 





d. BS% 




e. P-Index 



f. Mn-Index 





Figure 9. DCA of ginseng abundance and soil components. Crosses (+) indicate plot locations 
and cross size is proportional to ginseng abundance (a) or percentages of soil components (b-f). 



16 



Part I: HABITAT MODELING 

Results 

Canonical Correspondence Analysis CCCA) 

CCA of the canopy layer dominants against 1 1 environmental variables identified elevation, 
slope, and pH as having the most influence on plant species composition (Table 5). All three 
axes were significant according to a Monte Carlo permutation test (P < 0.05). 

Table 5. Ordination results for CCA of canopy layer dominants and environmental variables (ter 
Braak 1997). 





Axis 1 


Axis 2 


Axis 3 


Eigenvalues 


0.293 


0.240 


0.157 


Species environment correlations 


0.787 


0.745 


0.683 


Cumulative % variance 








of species data 


3.2 


5.8 


7.5 


Intraset correlations 








Elevation 


-0.765 


0.577 


-0.007 


Slope 


0.649 


0.397 


0.501 


pH 


-0.423 


-0.490 


0.742 



The ordination of validation plots shows the relative importance of pH, elevation, and percent 
slope (Figure 10). Plots in which ginseng was present typically had higher pH, lower elevation, 
and lower percent slope. In contrast, many outlier plots (e.g., in high elevations, or on steep 




• Ginseng Absent 
O Ginseng Present 



Figure 10. CCA plot ordination of dominant canopy species. Each circle represents a validation 
plot (n = 73). 



17 



Part I: HABITAT MODELING 

Results 



slopes) did not contain ginseng. 



Distribution Map 

A distribution map of currently known ginseng occurrence in GSMNP was produced using the 
52 positions and ginseng protection and validation data. The map identifies highly suitable, 
moderately suitable, and unsuitable ginseng habitat in the Park. Due to data sensitivity, the map 
is not included in this report. 

Abundance Estimate 

The model estimate for ginseng was produced from average habitat saturation and density for a 
refined portion of the study area (below the D 2 cutoff of 8.5 comprised of 136,519 pixels). The 
abundance estimate for the 8.5 cutoff was 51,195 plants. Because ginseng occurred beyond this 
cutoff in greater densities, but less frequently, an abundance estimate was also calculated for an 
extended portion of the study area {D 2 cutoff = 12.7). The higher cutoff yielded a total estimate 
of 212,559 ginseng plants. This cutoff was selected based upon the range of habitats in which 
ginseng and its associated plant species were predicted with reasonable accuracy. The predictive 
ability of the model was consistent for ginseng and the four associated species for all D 2 cutoff 
values tested. 



18 



Part I: HABITAT MODELING 
Discussion, Summary, and Recommendations 

DISCUSSION 

The modeling effort increased our understanding of ginseng habitat in GSMNP. Results indicate 
that ginseng occurs most often in rich cove forests (smaller D 2 values). However, ginseng 
occurred in a wider range of habitats than previously believed, and a greater abundance of ginseng 
was encountered in habitats with higher D 2 values. During model development, we focused on 
rich cove forests, but, because poachers target these areas, we expanded the range of potential 
habitats. Two ginseng populations (approximately 160 plants each) were encountered en-route to 
random plot locations. These populations were not included in the model validation, however 
similar habitat was described in plots where ginseng occurred. These habitats generally were 
small-scale, with a limited composition of cove forest species and D 2 values greater than the 
model cutoff 8.5. All ginseng occurrences had a D" less than 16.8. We describe this habitat 
range, 8.5 to 16.8, as moderately suitable in the distribution map and remainder of this discussion. 

Ginseng Habitat Description 

Associated Plant Species 

As expected, there was a positive correlation between the four associated plant species and 
ginseng occurrence. Sanguinaria, Cimicifuga, Adiantum, and Cyphpedium spp. consistently were 
found in similar habitats as ginseng. In contrast, there was weak negative correlation between 
ginseng abundance and Sanguinaria and Cimicifuga, and no correlation with Cyphpedium and 
Adiantum. This suggests that poachers may also use these species to locate ginseng, or, ginseng is 
currently concentrated in moderately suitable habitats with limited composition of associated 
species. Parallel analyses were performed for ginseng and the associated plant species in order to 
detect habitat characteristics and distribution patterns confounded by the effects of poaching. All 
data analyses for ginseng and its associated plant species, and tests to determine the model cutoff, 
suggest Sanguinaria, Cimicifuga, Adiantum, and Cyphpedium are effective surrogates for 
predicting ginseng and its habitat. 

Large-scale Patterns 

DCA grouped validation plots into three broad habitat types: high elevation forest, cove forest, 

and dry, ridge-top forest communities. This pattern has been reported in previous GSMNP 



19 



Part I: HABITAT MODELING 
Discussion, Summary, and Recommendations 

vegetation studies (Hicks 1980, Muller 1982, Parker and Parker 1986, Callaway et al. 1987) and 
demonstrates that the data set we used represents habitat types in the Park. However, the cove 
forest community type appears to include a range of habitats, and ginseng can occur in any of 
them. Within the cove forest group (Figures 7, 8, and 9), ginseng, its associated plant species, and 
high percentages of calcium, magnesium, and basic solvency, generally occurred together. 

CCA of dominant canopy species was consistent with earlier vegetation distribution studies in 
GSMNP, where elevation, slope, and pH were identified as determining factors of canopy 
composition (Cain 1931, Whittaker 1956, Hicks 1980). 

Plots where ginseng occurred generally possessed a dense, broadleaf sub-canopy and infrequent 
occurrence of evergreen species. In contrast, dense, species-poor stands of rosebay 
rhododendron {Rhododendron maximum) and/or Eastern hemlock (Tsuga canadensis), with a 
species-poor understory, were regularly encountered in the validation plots. 

Small-scale Patterns 

Larger populations of ginseng located during this study occurred in pockets of habitat that were 
too small (less than 900 m , or one pixel) to be detected by conventional GIS techniques. This 
limited our ability to incorporate these moderately suitable, small-scale habitats into the 
predictive habitat model. Ginseng's consistent occurrence in moderately suitable habitats in 
association with yellow buckeye (Aesculus flava), white basswood (Tilia americana var. 
heterophylla), and black locust (Robinia pseudoacacia) suggests strong plant species 
associations. The co-occurrence of herbaceous species, such as ginseng, with specific canopy 
species may be due to soil-enhancing mycorrhizal fungi-tree associations or other site-enhancing 
qualities (Newman and Reddell 1988). Earlier research in GSMNP has demonstrated that small- 
scale gradients, such as within-stand microtopography and soil gradients, determine herbaceous 
species distributions (microhabitats), but the overall distribution of habitats is determined by 
canopy associations and large-scale environmental gradients (Bratton 1976, Hicks 1980, 
Newman and Reddell 1988). 



20 



Part I: HABITAT MODELING 
Discussion, Summary, and Recommendations 

Microtopography and Soil Factors 

The results of all analyses of species occurrence and soil factors were consistent with earlier 
research (Bratton 1976, Hicks 1980) and suggest that soil composition influences ginseng 
distribution (Nault et al. 1998). Bratton (1976) identified a close correlation between herbaceous 
species' distributions and intra-stand soil heterogeneity. Patterns of pH and ginseng distribution 
suggest that moderately suitable habitats with high densities of evergreen sub-canopy species 
possess higher pH usually not reflected by the occurrence of these acid-tolerant species 
(requiring low pH). This suggests a moderating influence that creates a suitable microhabitat in 
otherwise unsuitable habitat. This requires further study due to the potential role microhabitats 
serve in maintaining and re-establishing ginseng populations. 

Abundance Estimate 

The estimate of ginseng abundance using the 8.5 D cutoff is 51,195 plants. An additional 
estimate of 212,559 plants is calculated using the D 2 cutoff 12.7, which includes moderately 
suitable habitats occupied by ginseng and its associated plant species. 



21 



Part I: HABITAT MODELING 
Discussion, Summary, and Recommendations 

SUMMARY 

There are several possible explanations for ginseng distribution in GSMNP including: local 
extirpation due to illegal harvest, land-use history, limited population re-establishment due to 
dispersal or physiological limitations, and/or opportunistic occupation of canopy gaps in 
otherwise unsuitable habitats. Contrary to our hypothesis, ginseng was sparse in optimal rich 
cove forests, and was found in greater numbers in moderately suitable habitats - habitats not 
targeted by poachers. This unexpected distribution pattern strongly suggests that poaching is a 
serious problem in GSMNP. 

The results of this study will assist in future habitat assessment and plant conservation efforts by: 

(1) identifying potentially suitable ginseng habitat to focus search efforts, especially when 
working across large landscapes, 

(2) locating additional populations to refine the habitat model and to better understand the 
ecology of ginseng, and 

(3) re-evaluating appropriate and sustainable levels of ginseng harvest, if any. 



22 



Part I: HABITAT MODELING 
Discussion, Summary, and Recommendations 

RECOMMENDATIONS 

During this two-year study, Park staff found only two populations of more than 160 plants. 
Further searches for viable populations are needed. Examination of aerial and infrared photos to 
locate marginal habitats and "pockets" should be used to direct on-the-ground searches. 

Associated plant species should be used to model ginseng habitat to circumvent the confounding 
effects of poaching. Roads, trails, and boundaries should be assessed for their role as poaching 
corridors. 

Demographic studies to compare GSMNP ginseng populations with populations in other areas of 
the eastern U.S. and Canada, may help us understand similarities and differences among 
populations. Such studies could provide crucial data needed to help conserve ginseng throughout 
its range. 

There are a number of other biotic factors that could affect ginseng distribution and, therefore, 
merit further study, including: 

(1) forest structure and structural/functional diversity in ginseng habitats, 

(2) exclusion of ginseng or its associated plant species by competitors or pathogens, 

(3) genetic variability within and between ginseng populations, and 

(4) seed dispersal mechanisms. 



23 



Part II: PROTECTION 
Introduction 

Part II: 
PROTECTION 

INTRODUCTION 

Illegal Harvest from GSMNP 

Park biologists and law enforcement rangers began keeping records of the amount of ginseng 
(number of roots and weight) illegally harvested from GSMNP in 1991 . In only the past nine 
years, GSMNP rangers have seized 10,515 ginseng roots from illegal harvesters (Table 6). In 
addition, rangers state they detect only a small percentage - only 1 to 3% - of people poaching 
ginseng within the Park, and actually apprehend fewer still. Rangers' duties at GSMNP are 
varied and only a fraction of their time can be spent in the backcountry focused on resource 
protection. 



Table 6. Summary of amount of illegally harvested ginseng (number of roots and undried 
weight) seized in Great Smoky Mountains National Park, 1991 - 1999. 



Year 


No. of roots seized 


Undried weight (g) 


1991 


846 


2,347.97 


1992 


399 


1,447.83 


1993 


3,301 


11,394.78 


1994 


1,723 


7,052.73 


1995 


1,846 


5,953.36 


1996 


1,797 


7,691.79 


1997 


52 


154.59 


1998 


365 


1,843.85 


1999 


186 


680.57 


Total 


10,515 roots 


38,567.47 




(35.38 dry lbs. @ 298 roots/lb.) 


(85.04 lbs. fresh) 



Despite this fact, from 1991 through 1999 rangers seized 85.04 pounds of fresh roots, equivalent 
to 35.28 dry pounds (Tables 6 and 7) (an average of 298 roots per dry pound from data collected 
from North Carolina dealers, by North Carolina Department of Agriculture, over the past nine 
years) (J. Corbin, NCDA; pers. comm.). 



24 



Part II: PROTECTION 
Introduction 

The current value of ginseng, based on 1999 figures, ranges from $475.00 per dry pound (value 
to digger) to $1,900.00 per dry pound (market value overseas, a four-fold increase over digger 
price) (Table 7) (M. Boyer, NCDA; J. Corbin, NCDA; and C. Robbins, TRAFFIC; pers. comm.). 

If a conservative 1% apprehension rate is applied, and the amount of ginseng apprehended is 
multiplied by 100, then approximately $1,330,056.00 (value to digger) to $5,320,224.00 (market 
value overseas) worth of ginseng roots have been illegally harvested from GSMNP in just the 
past 9 years. This figure is calculated from the average price per dry pound ($377) over the last 
nine years (Table 7). 

Table 7. Average number of roots per dry pound, and average price per dry pound paid to 
diggers, 1991 - 1999. 1 



Year Average number of roots/dry pound Average price /dry pound 

1991 295 $310.00 

1992 295 310.00 

1993 299 575.00 

1994 288 300.00 

1995 285 350.00 

1996 314 350.00 

1997 291 385.00 

1998 315 325.00 

1999 296 490.00 

Average 298 roots $377.00 

data gathered from North Carolina ginseng dealers (J. Corbin, NCDA; pers. comm.) 



Further, harvesting techniques are directly destructive to ginseng populations because the root is 
removed from the soil in its entirety leaving nothing behind to produce a new plant. Destruction 
to the habitat, and trampling and disturbance of neighboring plants occurs as well. Such 
substrate destruction, in itself, further impedes the recovery, if any, of any plants left behind. 
Poaching has become more sophisticated through the years. Often, several days are spent 
digging in remote off-trail areas. Poachers arrange drop-offs and pick-ups to avoid detection. 
They also cache roots for later pick-up. Recently it was observed that poachers were covering 
the holes created from plant removal to avoid post-harvesting detection. Tragically, in many 



25 



Part II: PROTECTION 
Introduction 

seized bags there are roots as young as one year up to 48 years of age. Although poachers appear 

to be most active in late summer, some are harvesting roots while the fruit/seeds of the ginseng 

plant are still green and on the plant, thus removing all reproductive potential from the 

population. Seizures by rangers have been made in early July - weeks before ginseng fruit/seeds 

ripen. If any plants were left in a poached population, they typically are not able to regenerate, 

or recover, before being poached again. In most cases, only a minimal seed reservoir remains in 

the soil. 

Ginseng in GSMNP 

Due to the biology of the ginseng plant and the nature of harvesting techniques, plants are unable 
to recover from harvest (natural history information on American ginseng is provided in Part I: 
Introduction). To reiterate, ginseng reproduces mainly by seed and rarely propagates by 
producing more than one rhizome. Plants must be older than five years to produce viable fruit, 
and the seeds require two winters of stratification before germinating. When combined, these 
factors result in slow rates of population growth (Lewis and Zenger 1982). 

Genetic analysis using Randomly Amplified Polymorphic DNA (RAPD) markers showed 
GSMNP ginseng to possess unique genetic integrity and may represent a distinct center of 
genetic diversity when compared to wild ginseng populations in Pennsylvania and Wisconsin (C. 
Boehm, University of Wisconsin, pers. comm.). To lose, or even reduce, populations of ginseng 
on a local scale will represent a significant loss for the species as a whole. 



26 



Part II: PROTECTION 
Methods and Materials 



METHODS and MATERIALS 



Protection Techniques 

GSMNP rangers and a plant specialist from North Carolina Department of Agriculture (NCDA) 
developed two marking techniques to deter poaching. The first technique involves applying a 
semi-permanent, powdered dye to the root of ginseng plants (by partially excavating a portion of 
the root surface for application then recovering with soil). The dye serves as an immediate 
deterrent and can be detected when the poacher sells collected roots to a dealer. The dye 
application method was used on roots of small, two- or three-pronged plants, or in combination 
with the following technique: the insertion of a stainless steel ribbon into the top portion of larger 
roots (usually three- and four-pronged plants). Once the ribbon is inserted under the surface of 
the root body, it can be detected with the use of a strong magnet or by x-ray. Roots were marked 
with a gauge of ribbon specific to one of the five ranger districts. This allows law enforcement 
officials to determine the source of seized roots within the Park and aids Park biologists in 
replanting the roots (if the roots are still viable) in or near the same general area. Any evidence 
of human activity at ginseng locations, such as digging, construction or maintenance of 
unofficial trails, and illegal camping were reported to district rangers. 

Data collection 

Areas to search in the field were identified through: 

(1) aerial photographs and USGS quadrangle maps (7.5 minute, 1:24,000 scale), 

(2) locations of extant GSMNP ginseng populations gathered from Park staff 
observations, 

(3) Park natural history records, 

(4) Park law enforcement observations on poaching activity, and 

(5) vegetation type (e.g., cove forest, northern hardwood forest) and topographic 
attributes such as slope, aspect, and elevation. 

These areas were located in the field using topographic features, compass bearings, and altimeter 
readings. A thorough search for ginseng was made when bloodroot (Sanguinaria canadensis), 
black cohosh (Cimicifuga spp.), maidenhair fern (Adiantum pedatum), and yellow lady's slipper 
(Cypripedium pubescens) or suitable habitat was present. 



27 



Part II: PROTECTION 
Methods and Materials 

At each site, ginseng plants were flagged, aged (by counting the annual bud scars on excavated 
rhizomes), and marked with one of the two techniques described above. Ginseng plants 
encountered within, or during navigation to, the validation plots (see Part I) were also marked. 
When a population was discovered, its location was pinpointed on a topographic map and. 
whenever possible, a digital position reading obtained using a GPS unit (PLGR + 96). 

Data collected at each ginseng location included dominant aspect, slope, and elevation (Data 
sheet, Appendix 6). Percent cover was estimated for canopy and the five dominant species were 
ranked. Information on reproductive status, age, and marking technique were recorded for each 
plant. The presence of one or all of the associated plant species was noted as well. 

Size Class and Age Distribution 

According to ginseng researchers (Lewis and Zenger 1982, Charron and Gagnon 1991), the 
structure of a ginseng population should contain mostly three-pronged plants. Four-pronged 
plants (typically the best reproducing individuals) are the least abundant. In order to determine 
the structure of each ginseng population, the number of prongs per plant was recorded. The ages 
of all three- and four-pronged plants were determined during the marking procedure. All roots 
excavated for marking were carefully aged without damage to the root or rhizome. Ages were 
recorded and graphed to show an overall age distribution for all plants found. 



28 



Part II: PROTECTION 
Results 



RESULTS 



Habitat Data 

The most common dominant canopy species associated with the protection sites were tuliptree 
{Lihodendron tulipifera), white basswood (Tilia americana var. heterophylla), and sugar maple 
{Acer saccharum). Also present were silverbell {Halesia tetraptera var. monticola), white ash 
(Fraxinus americana), yellow buckeye (Aesculus flava), and black locust (Robinia 
pseudoacacia). Aspects were to the north ranging from 354 °- 10°. Elevation range was 610 
meters (2,000 feet) up to 1,024 meters (3,360 feet). All sites had a canopy cover greater than 
75%. 

Associated Plant Species Occurrence 

Habitat modeling data showed positive correlations among the presence of ginseng and the four 
associated plant species {Sanguinaria, Cimicifuga, Adiantum, and Cypripedium) (Table 3, Part I). 
However, there were sites in which one or all four associated species were found, but ginseng 
plants were absent. 

Size Class Distribution 

A total of 501 plants was found in 46 locations during the 1998 protection effort. Eighty-two 
plants (16.4%) were marked by one of the two techniques and 97.5%. Forty-five percent of all 
plants were three-pronged and only 8.6% were four-pronged (Figure 1 1). 



29 



Part II: PROTECTION 
Results 



250 



200 



.£ 150 




100 



# of prongs 

Figure 11. Size class distribution of ginseng plants encountered in 1998 (n = 501). 

Age Distribution 

Eighty marked plants were aged. The results were consistent with Lewis and Zenger (1982). 
The ages of all ginseng plants encountered during the 1998 protection effort ranged from three to 
20 years, with an average age of seven years (Figure 12). 

18 
16 
14 
12 



a. 

■S 8 

6 

4 

2 





I 




I.I.I 



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 

age of plants (years) 

Figure 12. Age distribution of naturally occurring ginseng plants encountered during the 1998 
field season efforts. 



30 



Part II: PROTECTION 
Results 

Seized Roots and Poaching Evidence 

During this two-year project, GSMNP rangers seized a total of 354 ginseng roots from poachers. 
These roots were aged by Park biologists and then replanted. Ages of seized roots ranged from 
one to 47 years, with an average age of 15 years (Figure 13). Evidence of poaching, including 
obvious signs of digging and other human disturbance, were noted at one third of the 46 ginseng 
locations. 




' S 7" 



Figure 13. Age distribution of ginseng plants seized in 1997 and 1998. 



31 



Part II: PROTECTION 
Discussion 



DISCUSSION 



Much effort and staff resources were directed towards accomplishing all aspects of the study. In 
particular, the protection effort (marking plants in the field) employed six full-time biological 
science technicians, including more field technicians working off-trail on this and other natural 
resource projects. Despite the scope of the project and the involvement of seasoned technicians, 
fewer than 1,000 ginseng plants were encountered and no additional viable populations (172 
plants or more) (Nantel et al. 1996) were discovered. Many areas that were reported to have 
ginseng as recently as 1996 were found to have very few, if any, plants remaining in 1998. Over 
600 plants within one drainage, discovered by Park rangers in August 1997, were poached less 
than a year later. In another instance, a ginseng population was discovered by field technicians 
on one day, and was poached before noon the very next day. 

We were disheartened by the advanced ages of the roots seized in 1998 - the trend seems to be 
shifting to older plants. The oldest root was 47 years, and the mean age of all seized roots was 
16.1 years (SD = 7.7). Compared with data from past years, this is the highest mean age 
(although not statistically significant). In 1991, the mean age was 13.2 years, higher than the 
mean ages for 1992 through 1996 (range 8.6 to 11.1 years). 

It is evident from our study that an increase in protection efforts in GSMNP is critically 
important. Marking of ginseng roots by Park rangers and biologists will continue over the next 
several years, or until a reduction in poaching activity is seen. 



32 



LITERATURE CITED 

LITERATURE CITED 

Argus, G. W. and J. White. 1989. Panax quinquefolium L. in: G.W. Argus and C.J. Keddy 

(Eds.). Atlas of the rare vascular plants of Ontario. Part 3. Museum of Natural Sciences. 
National Museum of Canada, Ottawa. 

Beatty, S. W. 1984. Influence of microtopography and canopy species on spatial patterns of 
forest understory plants. Ecology, 65: 406 - 1419. 

Beers, T.W., P.E. Dress, and L.C. Wensel. 1966. Aspect transformation in site productivity 
research. Journal of Forestry 64: 691-692. 

Beven, K. J., and M. J. Kirby. 1979. A physically based, variable contributing area model of 
basin hydrology. Hydrol. Science Bull. 24: 43-69. 

Boetsch, J. R. and J. Rock. 1999 unpublished. Habitat modeling and conservation status of four 
vascular plants endemic to the southern Appalachian Mountains: National Park Service, 
Great Smoky Mountains National Park. 52 p. 

Bratton, S. P. 1976. Resource division in an understory herb community: responses to temporal 
and microtopographic gradients. The American Naturalist, 110 (974): 679-693. 

Cain, S. A. 1931. Ecological studies of the vegetation of the Great Smoky Mountains of North 
Carolina and Tennessee. I. Soil reaction and plant distribution. Botanical Gazette, 91 : 
22-41. 

Callaway, R. M., E. C. Clebsch and P. S. White. 1987. A multivariate analysis of forest 
communities in the western Great Smoky Mountains National Park. The American 
Midland Naturalist, 118: 107-120. 

Charron, D. and D. Gagnon. 1991. The Demography of Northern Populations of Panax 
quinquefolium (American Ginseng). Journal of Ecology, 79: 431-445. 

Golden, M. S. 1981. An integrated multivariate analysis of forest communities of the central 
Great Smoky Mountains. American Midland Naturalist, 106: 37-53. 

Halpin, P. N. 1995. A cross-scale analysis of environmental gradients and forest patterns in the 
Giant Sequoia - Mixed Conifer forest of the Sierra Nevada. Ph.D. dissertation, 
University of Virginia, Charlottesville. 

Hicks, D. J. 1980. Intrastand distribution patterns of southern Appalachian cove forest 
herbaceous species. American Midland Naturalist, 104 (2): 209-223. 

Hill, M. O. and H. G. Gauch. 1980. Detrended correspondence analysis, an improved ordination 
technique. Vegetatio, 42: 47-58. 



33 



LITERATURE CITED 

Hu, S. Y. 1976. The Genus Panax (Ginseng) in Chinese Medicine. Economic Botany, 30: 11- 
28. 

Lewis, W. H. and V. E. Zenger. 1982. Population Dynamics of the American Ginseng Panax 
quinquefolium (Araliaceae). American Journal of Botany, 69(9): 1483-1490. 

MacKenzie, M. D. 1993. The vegetation of Great Smoky Mountains National Park: past, 
present and future. Ph.D. Dissertation, University of Tennessee, Knoxville, TN. 

McCune, B., and M. J. Medford. 1995. PC-ORD. Multivariate Analysis of Ecological Data, 
Version 2.0. MjM Software Design, Gleneden Beach, Oregon, USA. 126 p. 

McNab, W. H. 1993. A topographic index to quantify the effect of mesoscale landform on site 
productivity. Canadian Journal of Forest Research 23 : 1100-1107. 

Miller, R. 1986. Predicting rare plant distribution patterns in the southern Appalachians of the 
southeastern U.SA. Journal of Biogeography 13: 293-311. 

Muller, R. N. 1982. Vegetation patterns in the mixed mesophytic forest of eastern Kentucky. 
Ecology, 63(6): 1901-1917. 

Nantel, P., D. Gagnon and A. Nault. 1996. Population viability analysis of American ginseng 

and Wild leek harvested in stochastic environments. Conservation Biology, 10: 608-620. 

Nault, A. 1997. Le situation du ginseng a cinq folioles (Panax quinquefolius L.) au Quebec. 
Gouvernment due Quebec, Ministere de l'Environement et de la faune, Direction de la 
conservation et du patrimoine ecologique, Quebec, 43 p. 

Nault, A., D. Gagnon, D. White, and G. Argus. 1998 unpublished. Conservation of Ginseng in 
Ontario. Ministry of Natural Resources, Science and Technology Report 1997/1998. 19 
P- 

Newman, E. I. and P. Reddell. 1988. Relationship between mycorrhizal infection and diversity 
in vegetation: evidence from the Great Smoky Mountains. Functional Ecology, 2: 259- 
262. 

Parker, A. J. 1982. The topographic relative moisture index: an approach to soil-moisture 
assessment in mountain terrain. Phys. Geogr. 3(2): 160-168. 

Parker, A. J. and K. C. Parker. 1986. A comparison of ordination techniques for interpreting 
forest composition in the Southern Appalachians. Southeastern Geographer, 26(1): 12- 
24. 

Rao, C. R. 1952. Statistical methods in biometric research. J. Wiley and Sons, New York, N.Y. 
390 p. 



34 



LITERATURE CITED 

Robbins, C. S. 1998. American Ginseng: the Root of North America's Medicinal Trade. Report 
by TRAFFIC North America, Washington, D. C. 94 p. 

Schafale, M. P., and A. S. Weakley. 1990. Classification of the natural communities of North 
Carolina: third approximation. North Carolina Heritage Program. 325 p. 

ter Braak, C. J. F. 1986. Canonical correspondence analysis: a new eigenvector technique for 
multivariate direct gradient analysis. Ecology 67:1 167-1 179. 

Whittaker, R. H. 1956. Vegetation of the Great Smoky Mountains. Ecological Monographs, 
26: 1-80. 

Wilds, S. P. 1996. Gradient analysis of the distribution of flowering dogwood (Cornus florida 
L.) and dogwood anthracnose (Discula destructive! Redlin.) in western Great Smoky 
Mountains National Park. M.S. Thesis, Univ. of North Carolina, Chapel Hill. 151p. 

Wolock, D. M. 1993. Simulating the variable-source-area concept of streamflow generation 
with the watershed model TOPMODEL. U.S. Geol. Surv. Water Resour. Invest. Rep. 
93-4124. 

Wolock, D. M. and G. J. McCabe, Jr. 1995. Comparison of single and multiple flow direction 
algorithms for computing topographic parameters in TOPMODEL. Water Resources 
Research 31(5): 1315-1324 



35 



APPENDICES 



APPENDICES 



36 



APPENDICES 



Appendix 1. GIS variables used to calculate Mahalanobis distance values (D 2 ) in GSMNP. 

Source 



Name 



Description 



Value Range 



Aspect 



Elevation 



Landform 
Index 

Planiform 



Aspect transformed using: 
Beers = 1 + cos (45-aspect) 

Elevation (meters) 



Index of meso-scale topographic 
protection 

Slope curvature in horizontal 
plane 



Profile Slope curvature in vertical plane 

Curvature 



Relative Slope Relative slope position (percent) 
Position 



to 2.0 



914-2,027 



-4.026 to 
93.587 

8.802 to 8.838 



-10.906 to 
8.639 



Oto 100 



Beers et al. 1966 



USGS digital 
elevation model (DEM) 

McNab 1993 



calculated from 
elevation with the 
curvature command (grid) 

calculated from 
elevation with the 
curvature command (grid) 

Wilds 1996 



Slope 



Solar 
Insolation 



Shannon- 

Weiner 

Index 



Slope steepness (degrees) 



Index of exposure to sunlight; 
approximated for the solar 
equinox; considers both nearby 
and distant shadowing factors 

Shannon- Weiner index of 
topographic complexity 



0.0 to 64.467 



1 to 227 



17 to 35 



calculated from 
elevation with the 
curvature command (grid) 

calculated from 
elevation with the 
hillshade command (grid) 



Miller 1986 



Topographic 
Convergence 
Index 



Topographic 
Relative 
Moisture 
Index 



Simulates the flow accumulation 1 7 to 1 82 

of water; TCI = ln(A7 tan B), 
where A is drained surface area 
and B is drained surface slope 

Index of moisture considering the to 85 

effects of slope position, aspect 
and elevation 



Beven and Kirkby, 
1979; Wolock 1993; 
Wolock and McCabe 
1995; Halpin 1995 

Parker 1982 



Terrain Shape Index of micro-scale 
Index topographic sheltering 



1 69 to 7 1 



McNab 1989 



37 



Appendix 1 continued 

Vegetation 



Vegetation type (9 forest types, 
heath bald, grassy bald, grape 
thicket, or treeless) 



APPENDICES 



categorical 
(1 to 13) 



McKenzie 1993 



a All variables were continuous and had base resolution of 30 m except for vegetation, which was a nominal variable. 

Vegetation had a base resolution of 90 m 

b Range values were for the study area, defined as GSMNP above 914 m. 



38 



APPENDICES 



Appendix 2. Ginseng model validation data sheet. 

Date: Surveyors: JRB JHR HH EC JT KC BY TM Ginseng Model Validation, 3d 

Plot ID #: Location: 

Coordinates: e n +/- m WP# GPS? Y N Quad: 



Elevation (ft): Aspect: Var: high mod low Slope (deg): Var: high mod low 

Terrain shape index (15 m): aspect +45 +90 +135 +180 +225 +270 

+315 

Microtopography: exaggerated well-developed modest non-existent Horizontal curvature (30 m): 

concave straight convex 

Slope position: toe low mid high crest Vertical curvature (30m): concave straight convex 

Surface water: inside plot <50m >50m Hydrology: permanent drainage intermittent drainage seep 

boulderfield upland 

Dominant stratum height: 30+m 20-30 10-20 5-10 0.5-5 <0.5m Gaps: numerous, 

influential few, nominal none 

Dominant stratum cover (%): >60 >25 >10 <10 Composition: evergreen (>75% of cover) deciduous 

(>75% of cover) mixed 

Cover (nearest 10%) Dominant species, in order (! principal, = equivalent dominance, ~ negligible) 

Canopy 20 40 60 80 100 Leaf litter: 

Sub-canopy 20 40 60 80 100 Evergreen: 

Herb 20 40 60 80 100 Nonvasc: 

Y N Sanguinaria Y N Cimicifuga spp. Y N Adiantum Y N Cypripedium pubescens 

Rock cover: Type: bedrock boulders talus cobbles gravel Old-growth? Y N ?Soil sample 

collected? Y N 

Disturbance: none BWA beech scale rooting exotics homesite Degree: Slight Mod Severe 

Standnotes: 

Panax Panax poached? Y N 

Present in plot (31.6m 2 )? Y N 

Abundance class for plot (below): Inf: 4P 3P 2P IP 

Fertility class for plot (below): Fert: 4P 3P 2P 1 P 

Are plants generally healthy? Y N Panax health: Anthracnose Mildew Alternaria 

Notes: 



Abundance: 1: < 10 2:10-24 3:25-49 4:50-99 5:100-199 6:200-499 7:500-999 8:1000- 
10,000 9:10,000 + 

Fertility: 1: < 10% 2: 10 - 40% 3: 40 - 60% 4: 60 - 90% 5: >90% Subplot: 5: (.316m) 2 4: (lm) 2 3: 
(3.16m) 2 2: (15.8m) 2 I: plot 

Canopy Species Coll. Sub-canopy Species Coll. Herb Species Coll. 



39 



Appendix 3. Plant Species List For DCA Plots 



APPENDICES 



A. Dry Ridgetop Plots 

Acer rubrum 
Carya alba 
Castanea dentata 
Chimaphila maculata 
Cornus florida 
Dryopteris intermedia 
Galax aphylla 
Gaylussacia ursina 
Huperzia sp. 
Kalmia latifolia 
Leucothoe editorum 
Liriodendron tulipifera 
Medeola virginiana 
Oxydendron arboreum 
Pinus echinata 
Pinus rigida 
Pinus strobus 
Pyrularia pubera 
Quercus coccinea 
Quercus montana 
Quercus velutina 
Rhododendron maximum 
Thelypteris noveboracensis 
Tsuga canadensis 
Uvular ia pudica 
Viburnum acerifolium 



B. Cove Hardwood Plots 

Acer rubrum 

Adiantum pedatum 

Arisaema triphyllum 

Betula lenta 

Carya alba 

Castanea dentata 

Cimicifuga racemosa 

Clintonia umbellulata 

Cornus florida 

Dryopteris intermedia 

Halesia tetraptera var. monticola 

Huperzia lucidula 

Impatiens sp. 

Laportea canadensis 

Leucothoe fontanesiana 

Liriodendron tulipifera 

Lysimachia quadrifolia 

Medeola virginiana 

Monarda didyma 

Panax quinquefolius 

Parthenocissus quinquefolius 

Pinus strobus 

Pyrularia pubera 

Quercus velutina 

Rhododendron maximum 

Rudbeckia lacinata 

Sanguinaria canadensis 

Sanicula sp. 

Thelypteris noveboracensis 

Tiarella cordifolia 

Tilia americana var. heterophylla 

Tsuga canadensis 

Uvularia pudica 

Viburnum acerifolium 

Viola hastata 



C. High Elevation Plots 

Abies fraseri 
Acer rubrum 
Acer spicatum 
Athyrium felix-femina 

spp. asplenioides 
Betula alleghaniensis 
Clintonia borealis 
Dryopteris intermedia 
Huperzia lucidula 
Ilex montana 
Impatiens sp 
Laportea canadensis 
Oxalis montana 
Picea rubens 
Prunus pensylvanica 
Rhododendron catawbiense 
Rubus canadensis 
Sambucus canadensis 
Sambucus pubens 
Tiarella cordifolia 
Viburnum lantanoides 



40 



APPENDICES 



Appendix 4. Ginseng protection data sheet. 



Date: 

Location: 



Surveyors: JRB JHR HH EC JT BY TM 



Ranger District: Cades Cove Little River East Oconaluftee Lake 

Coordinates: e n +/- m GPS? Y N WP# 

Quad: 



Elevation (ft): 



Aspect: 



Slope (deg): 



Slope position: toe slope low slope mid slope high slope crest Soil sample collected? Y 

N 

Canopy Composition: evergreen (>75% of cover) deciduous (>75% of cover) mixed 

Cover (nearest 10%) Dominant species, in order (Iprincipal species, =equivalent dominance, -negligible) 
Canopy 20 40 60 80 100 

Sab-canopy 20 40 60 80 100 

Herb 20 40 60 80 100 

Disturbance: none poaching beech scale hog rooting exotic spp homesite 

_ Sanguinaria canadensis 

Cimicifuga spp. 
_ Adiantum pedatum 

Cypripedium pubescens 

Notes: 





1 pr 


2 pr 


3 pr 


4 pr 


# fertile 










# infertile 










Total 













1 pr 


2 pr 


3 pr 


4 pr 


# marked with 
dye 










# marked with 
steel ribbons 










ages 

of marked plants 











41